09-19-2025, 05:40 PM
The effect of artificial intelligence (AI) in healthcare, in particular with respect to determinants of health or illness, is primarily one of greater depth and certainty in the analysis of population data. The ethical issues inherent in aggregation and use of massive health data warehouses consist primarily of the risks involved in the potentially expedient means to achieve the value promised through application of AI, along with the usual barely legal methods employed in commercial exploitation of such. Almost any behavioral or social determinant of health can be studied using artificial intelligence provided sufficient data with relevant content is available. This includes the value of health data streams directly from sensors worn or implanted in the patient themselves, the advantage of AI being its ability to learn from the vast quantities of detailed data which are thereby made available. For the purposes of this paper, the focus will be on the utility of health data warehouses for studying determinants of health in populations.
Population studies of the behavioral determinants of health, such as studies of nutritional and exercise habits and practices along with recreational drug use and the ethically distinct category of individual liberty in drug use, that promoting health, can all be potentially studied through application of deep learning AI to health data warehouses. Social determinants of health are intertwined with individual behavioral ones primarily to the extent that they influence individual behavior, although the built environment’s contributions to health outcomes cannot be ignored. Obvious examples include the presence of lead in water plumbing and the prevalence of car exhaust caused smog prior to the legal requirement of catalytic converters on all gasoline combustion engines. One less obvious example currently under study is that of the presence of microplastics in the environment causing health effects through their tendency to enter through ingestion. Determinants of health which can be studied with AI applied to data warehouses are the focus of this paper, and ultimately this includes almost all of them. The primary drivers of ill health therefore should be selected, such including exercise, nutrition, drug use, and cultural membership or beliefs affecting selection or rejection of health protecting or promoting behaviors generally.
The Health Belief Model provides a framework for study of both cultural influences and beliefs respecting adoption of health promoting behaviors and cessation of unhealthy ones. It’s basic premise is that the primary motivators of behavior change are beliefs respecting the risks inherent in particular behaviors and those respecting the benefits achievable with adoption of others (Nash et al., 2021). The determinant in this case, influencing all others, is belief.
One type of analysis done by social media platforms that addresses beliefs is called sentiment analysis. According to Wikipedia (Sentiment Analysis), this “is the use of natural language processing, text analysis, computational linguistics, and biometrics to systematically identify, extract, quantify, and study affective states and subjective information.” With the deepening power of AI applied to advertising on such platforms, particular product advertisements could be shown to particular users with a high likelihood of believing that that product would be good to acquire. Applied to health behavior modification, particular messages could just as easily be targeted to those most susceptible to receiving and acting on them based on the beliefs that can be inferred form the information available on them. AI applied to health data warehouses in combination with content of social networking sites would be particularly effective in custom tailoring messages that motivate healthy behaviors or inhibit unhealthy ones. Social media platforms can provide escape from physical cultural membership constraints and an avenue to new social experiences, and there is no reason to waste the beneficial influence these new social interactions can have to reinforce beliefs that support healthy behaviors.
Positive messaging respecting exercise can come from authoritative sources such as the American College of Sports Medicine, which helped craft the exercise promoting health messaging of the United Stated government which started in the early 1990’s (ACSM, 2025). The dramatic decline in the incidence of chronic diseases caused by sedentary lifestyle induced thrombosis, or abnormal blood coagulation, occurring during the period since then can be no coincidence, as such authority supported the growth of the inexpensive popular gym industry exemplified by Planet Fitness.
Health equity is frequently frustrated, however, by nutritional behaviors, those of the lower income demographics showing an increasing trend toward lack of attention and discipline in this behavioral determinant of health (Ong et al., 2024). The prevalence of high energy, low nutritional value foods and the busy, active work life of the low income family create a reliance on convenience food at supermarkets and fast food restaurants. The recent enactment of requirements for nutritional labeling of fast foot items is a form of messaging which simply makes selections more competent, providing information to the consumer and incentive to the producer to improve the nutritional value of their products. Determining the effectiveness of such information availability can be a subject of study using AI applied to health data warehouses provided sales information, including historic, can be obtained from fast food restaurant chains. Dining behaviors can also be inferred from social media content, so with sufficient breadth of sources AI can help here also. This one example demonstrates the power of providing information needed for competent decision making, which is also the primary value to populations of the availability of inexpensive digital technology, something no software developer could miss.
Another impediment to health equity is the quality and types of recreational drugs available to lower income individuals and communities combined with their general lack of education respecting their short and long term health effects. Of the domains related to social determinants of health, the two most relevant to this determinant are education access and quality and social and community context. Both of these can be studied using AI applied to the current health data warehouses, which generally include social networking derived information also, in particular if one accounts for the popular authority of entertainers who are perceived as free from the usual social constraints which limit opportunities to lower socioeconomic strata generally. The social context provides readily available escapes and other solutions to social stratification with education being the hard road that few are willing to travel for the duration required. Messages reinforcing positive educational behaviors can be designed and tailored using application of AI to the vast quantities of data already available, and with the emerging ability to re-identify social networking sources, can be targeted for effect. The potential for re-identification can be seen as entirely ethically negative except for the rare exception which this application would be an example of, which of course must be carefully regulated or government controlled to have a constructive effect. One need only study the aggressive frauds of the cigarette industry as their campaign to preserve their marketing efforts was steadily shut down through regulation to understand the potential for evil inherent in such targeting.
The determinants of health which AI can be effective with respect to are numerous and diverse since it can be used to study any and produce competent analysis usable for effective policy designs as well as commercial IT product design. The ethics of application of AI to achieve the benefits of knowing what will motivate people to actually change their behavior, for instance, is a complex question fought with risks of various kinds. To what extent can we use AI to re-identify a subject for messaging or other interventions and what level of risk to the individual or others would justify it? Can we ignore the power of AI in identification of individuals at risk or a danger to others and therefore fail to implement the AI needed for re-identification? The ethics of studying exercise and nutrition attitudes and practices of particular demographics, among other valuable application, cannot be denied, and such study is ethically sound, although it does require a risky levels of health and social networking data integration into vast, high value warehouses which under currently law can be commercially exploited, sending potentially sensitive data to entities with unknown ethical standards. The risk needs mitigation with governance standards and competent regulation, making possible reasonably safe realization of the potential of AI to analyze and reveal the details of various population determinants of health.
References:
McCarron, T.L., Noseworthy, T., Moffat, K., Wilkinson, G., Zelinsky, S., W, D., Hassay, D.,
Lorenzetti, D. L., & Marlett, N.J. (2019), Understanding the motivations of patients:
A co‐designed project to understand the factors behind patient engagement,
Health Expectations, August, 2019.
Nash, D.B., Skoufalos, A., Fabius, R.J., Oglesby, W.H. (2021), Population Health, Creating
a Culture of Wellness, Jones and Bartlett Learning, LLC.
Ong, J. C. L., Seng, B. J. J., Law, J. Z. F., Low, L. L., Kwa, A. L. H., Giacomini, K. M., &
Ting, D. S. W. (2024). Artificial intelligence, ChatGPT, and other large language
models for social determinants of health: Current state and future directions.
Cell Reports. Medicine, 5(1), Article 101356. https://doi.org/10.1016/j.xcrm.2023.101356.
Thompson, W. (senior editor) (2010, 2025). ACSM’w Guidelines for Exercise Testing and
Prescription, 8th and 12th editions. Walters Kluwer, 2010 and 2025 respectively.
Population studies of the behavioral determinants of health, such as studies of nutritional and exercise habits and practices along with recreational drug use and the ethically distinct category of individual liberty in drug use, that promoting health, can all be potentially studied through application of deep learning AI to health data warehouses. Social determinants of health are intertwined with individual behavioral ones primarily to the extent that they influence individual behavior, although the built environment’s contributions to health outcomes cannot be ignored. Obvious examples include the presence of lead in water plumbing and the prevalence of car exhaust caused smog prior to the legal requirement of catalytic converters on all gasoline combustion engines. One less obvious example currently under study is that of the presence of microplastics in the environment causing health effects through their tendency to enter through ingestion. Determinants of health which can be studied with AI applied to data warehouses are the focus of this paper, and ultimately this includes almost all of them. The primary drivers of ill health therefore should be selected, such including exercise, nutrition, drug use, and cultural membership or beliefs affecting selection or rejection of health protecting or promoting behaviors generally.
The Health Belief Model provides a framework for study of both cultural influences and beliefs respecting adoption of health promoting behaviors and cessation of unhealthy ones. It’s basic premise is that the primary motivators of behavior change are beliefs respecting the risks inherent in particular behaviors and those respecting the benefits achievable with adoption of others (Nash et al., 2021). The determinant in this case, influencing all others, is belief.
One type of analysis done by social media platforms that addresses beliefs is called sentiment analysis. According to Wikipedia (Sentiment Analysis), this “is the use of natural language processing, text analysis, computational linguistics, and biometrics to systematically identify, extract, quantify, and study affective states and subjective information.” With the deepening power of AI applied to advertising on such platforms, particular product advertisements could be shown to particular users with a high likelihood of believing that that product would be good to acquire. Applied to health behavior modification, particular messages could just as easily be targeted to those most susceptible to receiving and acting on them based on the beliefs that can be inferred form the information available on them. AI applied to health data warehouses in combination with content of social networking sites would be particularly effective in custom tailoring messages that motivate healthy behaviors or inhibit unhealthy ones. Social media platforms can provide escape from physical cultural membership constraints and an avenue to new social experiences, and there is no reason to waste the beneficial influence these new social interactions can have to reinforce beliefs that support healthy behaviors.
Positive messaging respecting exercise can come from authoritative sources such as the American College of Sports Medicine, which helped craft the exercise promoting health messaging of the United Stated government which started in the early 1990’s (ACSM, 2025). The dramatic decline in the incidence of chronic diseases caused by sedentary lifestyle induced thrombosis, or abnormal blood coagulation, occurring during the period since then can be no coincidence, as such authority supported the growth of the inexpensive popular gym industry exemplified by Planet Fitness.
Health equity is frequently frustrated, however, by nutritional behaviors, those of the lower income demographics showing an increasing trend toward lack of attention and discipline in this behavioral determinant of health (Ong et al., 2024). The prevalence of high energy, low nutritional value foods and the busy, active work life of the low income family create a reliance on convenience food at supermarkets and fast food restaurants. The recent enactment of requirements for nutritional labeling of fast foot items is a form of messaging which simply makes selections more competent, providing information to the consumer and incentive to the producer to improve the nutritional value of their products. Determining the effectiveness of such information availability can be a subject of study using AI applied to health data warehouses provided sales information, including historic, can be obtained from fast food restaurant chains. Dining behaviors can also be inferred from social media content, so with sufficient breadth of sources AI can help here also. This one example demonstrates the power of providing information needed for competent decision making, which is also the primary value to populations of the availability of inexpensive digital technology, something no software developer could miss.
Another impediment to health equity is the quality and types of recreational drugs available to lower income individuals and communities combined with their general lack of education respecting their short and long term health effects. Of the domains related to social determinants of health, the two most relevant to this determinant are education access and quality and social and community context. Both of these can be studied using AI applied to the current health data warehouses, which generally include social networking derived information also, in particular if one accounts for the popular authority of entertainers who are perceived as free from the usual social constraints which limit opportunities to lower socioeconomic strata generally. The social context provides readily available escapes and other solutions to social stratification with education being the hard road that few are willing to travel for the duration required. Messages reinforcing positive educational behaviors can be designed and tailored using application of AI to the vast quantities of data already available, and with the emerging ability to re-identify social networking sources, can be targeted for effect. The potential for re-identification can be seen as entirely ethically negative except for the rare exception which this application would be an example of, which of course must be carefully regulated or government controlled to have a constructive effect. One need only study the aggressive frauds of the cigarette industry as their campaign to preserve their marketing efforts was steadily shut down through regulation to understand the potential for evil inherent in such targeting.
The determinants of health which AI can be effective with respect to are numerous and diverse since it can be used to study any and produce competent analysis usable for effective policy designs as well as commercial IT product design. The ethics of application of AI to achieve the benefits of knowing what will motivate people to actually change their behavior, for instance, is a complex question fought with risks of various kinds. To what extent can we use AI to re-identify a subject for messaging or other interventions and what level of risk to the individual or others would justify it? Can we ignore the power of AI in identification of individuals at risk or a danger to others and therefore fail to implement the AI needed for re-identification? The ethics of studying exercise and nutrition attitudes and practices of particular demographics, among other valuable application, cannot be denied, and such study is ethically sound, although it does require a risky levels of health and social networking data integration into vast, high value warehouses which under currently law can be commercially exploited, sending potentially sensitive data to entities with unknown ethical standards. The risk needs mitigation with governance standards and competent regulation, making possible reasonably safe realization of the potential of AI to analyze and reveal the details of various population determinants of health.
References:
McCarron, T.L., Noseworthy, T., Moffat, K., Wilkinson, G., Zelinsky, S., W, D., Hassay, D.,
Lorenzetti, D. L., & Marlett, N.J. (2019), Understanding the motivations of patients:
A co‐designed project to understand the factors behind patient engagement,
Health Expectations, August, 2019.
Nash, D.B., Skoufalos, A., Fabius, R.J., Oglesby, W.H. (2021), Population Health, Creating
a Culture of Wellness, Jones and Bartlett Learning, LLC.
Ong, J. C. L., Seng, B. J. J., Law, J. Z. F., Low, L. L., Kwa, A. L. H., Giacomini, K. M., &
Ting, D. S. W. (2024). Artificial intelligence, ChatGPT, and other large language
models for social determinants of health: Current state and future directions.
Cell Reports. Medicine, 5(1), Article 101356. https://doi.org/10.1016/j.xcrm.2023.101356.
Thompson, W. (senior editor) (2010, 2025). ACSM’w Guidelines for Exercise Testing and
Prescription, 8th and 12th editions. Walters Kluwer, 2010 and 2025 respectively.